Functionality-driven food formulation – Reducing the environmental impact using machine learning
Food products like vegan yoghurts or ready-to-eat soups are formulated with standardized and pure ingredients. The production of these ingredients is very intensive in many resources like water, energy, and raw materials. On top of that, splitting agro-materials into standardized ingredients and subsequently combining them again into food products is not efficient. Novel milder methods have recently been developed that produce ‘mildly processed’ ingredients which have a lower environmental impact and can deliver similar properties as the highly refined ingredients. Yet, these milder refined ingredients are different and not standardized. This complicates the supply chain and food product formulation. This study is devoted to making the formulation of foods using milder refined ingredients less complex with algorithms from the field of machine learning. The study showed amongst other things, that foods can be optimized in terms of nutritional profile, functional properties, yielding a very substantial reduction in the environmental impact.